The Federation Council has approved a law that, starting in 2027, will allow patenting technical solutions based on machine learning algorithms and models. Amendments are being made to the Civil Code. The document was prepared by senators and deputies led by Federation Council Speaker Valentina Matvienko.
Currently, patent law in Russia is focused on tangible objects, which complicates the protection of software solutions. The new law will allow patenting solutions at the level of technical operating principles, and not just at the code level. It is expected that the changes will be especially important for developers of industrial software and robotics, where the key value is not just the program text, but the technical architecture. In the future, it is planned to clarify the criteria for patentability of algorithms and machine learning models.
The law will come into force on January 1, 2027. This should increase the investment attractiveness of Russian IT companies, as well as lead to an increase in the number of IT patents. As noted in the Federation Council, the law will allow domestic IT companies to emerge from the “gray zone” and guarantee their barrier-free entry into the intellectual property market.
How algorithms are protected abroad
In foreign practice, approaches to patenting algorithms and AI models vary. In the USA, patentability is determined by the Alice/Mayo test – it is necessary to show that the algorithm provides a technological improvement to the system. In Europe, the criterion of technical character applies: the algorithm must solve a specific technical problem. In China, it is required that the solution interacts with hardware or provides a measurable technical effect. In the UK, after a recent Supreme Court decision, the approach has become closer to the European one. At the same time, all countries agree on one thing: only a human can be an inventor; AI systems cannot be authors of patents.